Predicting Operator Workload from Oculometric Data in High-Demand Environments: A Case Study with MATB-II

In safety-critical and high-pressure environments, professionals frequently encounter elevated cognitive states, including acute stress and mental workload (MWL), which can ultimately lead to burnout. MWL is defined as the ratio between the available cognitive resources and the demands of a given task, and can be assessed through subjective self-reports, behavioural analysis, or physiological signal monitoring. Among these methods, physiological monitoring stands out as the most promising approach due to its independence from specific tasks and its capacity for real-time application. This study aims to develop a non-contact system to estimate MWL levels based solely on ocular signals, which can be captured using wearable devices such as smart glasses or remote cameras. A cohort of 28 participants engaged in the Multi- Attribute Task Battery II (MATB-II) test, designed to induce cognitive workload through multitasking, including visuomotor coordination, auditory reflex, logical reasoning, and visual reflex. Additionally, a secondary arithmetic task was incorporated to further explore varying levels of workload. Ocular features associated with each test phase and the participant’s personal MWL evaluation were extracted, normalized, and used in a machine learning pipeline to predict MWL states. The results demonstrate reliable prediction, with an F1-score macro of 0.77, successfully distinguishing between rest, low, moderate and high MWL states.